### Answer --- This is a transcription accuracy issue where the speech-to-text model may mishear spelled letters. Here are some approaches to improve this: 1. Use Boosted Keywords Add common names or spelling patterns to bias the transcriber model so those words are more likely to get transcribed correctly. [(1) ]() 2. Switch Transcription Mode to Accuracy Change the

stt_mode

setting to

accurate

instead of

fast

to prioritize transcription accuracy over latency. , [(2) ]() [(3) ]() , [(4) ]() , [(5) ]() 3. Adjust Denoising Mode Play with the denoising mode setting to see if background noise is impacting transcription quality. [(1) ]() , [(2) ]() [(3) ]() , [(4) ]() , [(5) ]() 4. Prompt Engineering In your prompt, explicitly instruct the agent that when a user spells their name letter-by-letter, the spelled version should be treated as the source of truth—and if the transcription is unclear, have the agent ask again. [(2) ]() The core challenge is that the LLM receives the transcribed text, so if the transcription itself is incorrect, the agent will work with that incorrect data. Combining boosted keywords with accuracy mode and clear prompt instructions should help. ---

​ To learn more, see the sources I considered: [[1] Increase transcription accuracy](https://docs.retellai.com/reliability/wrong-transcript) [[2] Agent issue](https://discord.com/channels/1186424376300679209/1433446566429200486) [[3] Get Voice Agent](https://docs.retellai.com/api-references/get-agent) [[4] Update Voice Agent](https://docs.retellai.com/api-references/update-agent) [[5] Create Voice Agent](https://docs.retellai.com/api-references/create-agent) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.